Research map / Detection and segmentation / Open-set and few-shot detection
Learning from Rich Semantics and Coarse Locations for Long-tailed Object Detection
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add CLIP-derived soft semantic supervision from extra classification data | DINO with R50 on LVIS val v1.0 · LVIS val v1.0 | AP: 32.2 -> 35.0 | Helped |
| Add CLIP-derived soft semantic supervision from extra classification data | DINO with R50 on LVIS val v1.0 · LVIS val v1.0 | AP_r: 24.1 -> 30.4 | Helped |
| Use RichSem instead of Detic's extra-data training | DINO with R50 on LVIS val v1.0 · LVIS val v1.0 | AP_r: 26.9 -> 30.4 | Helped |
| Initialize the detector backbone with CLIP (ablation) | DINO with R50 on LVIS val v1.0 · LVIS val v1.0 | AP: 32.2 -> 30.6 | Hurt |
| Share parameters between semantic and classification heads (ablation) | DINO with R50 on LVIS val v1.0 · LVIS val v1.0 | AP: 32.2 -> 31.1 | Hurt |
| Add an independent semantic branch (ablation) | DINO with R50 on LVIS val v1.0 · LVIS val v1.0 | AP_r: 24.1 -> 30.4 | Helped |
| Add soft semantics loss for INet-LVIS extra data (ablation) | DINO with R50 on LVIS val v1.0 · LVIS val v1.0 | AP: 33.8 -> 35.0 | Helped |
| Add soft semantics loss for INet-Unl extra data (ablation) | DINO with R50 on LVIS val v1.0 · LVIS val v1.0 | AP_r: 23.7 -> 28.6 | Helped |
| Use unlabeled ImageNet-21k as extra data (ablation) | DINO with R50 on LVIS val v1.0 · LVIS val v1.0 | AP: 32.2 -> 34.7 | Helped |
| Use Object365 boxes without class labels as extra data (ablation) | DINO with R50 on LVIS val v1.0 · LVIS val v1.0 | AP: 32.2 -> 33.0 | Helped |
| Replace CLIP RN50 with the stronger RN50×4 teacher (ablation) | RichSem with R50 on LVIS val v1.0 · LVIS val v1.0 | AP: 35.0 -> 36.0 | Helped |
| Increase λ_soft from 0 to 1.0 without extra data (ablation) | DINO with R50 on LVIS val v1.0 · LVIS val v1.0 | AP_r: 24.1 -> 27.0 | Helped |
Closest papers by meaning
- SimLTD: Simple Supervised and Semi-Supervised Long-Tailed Object Detection (CVPR 2025 · similarity 0.91)
- Search and Detect: Training-Free Long Tail Object Detection via Web-Image Retrieval (CVPR 2025 · similarity 0.88)
- EAT: Towards Long-Tailed Out-of-Distribution Detection (AAAI 2024 · similarity 0.87)
- Is Complex Training Necessary for Long-Tailed OOD Detection? A Re-think from Feature Geometry (NeurIPS 2026 · similarity 0.87)
- Long-tailed Object Detection Pretraining: Dynamic Rebalancing Contrastive Learning with Dual Reconstruction (NeurIPS 2024 · similarity 0.86)
- Feature Fusion from Head to Tail for Long-Tailed Visual Recognition (AAAI 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.